论文部分内容阅读
针对蒙古语语音识别系统急待设计出相应问题集的现状,讨论了蒙古语问题集设计中的原则、音子集的选择、音素表的完善、复合元音的归类、松与紧概念的应用等关键问题的解决方案,给出了一个蒙古语标准音问题集.多次对比实验结果表明,该蒙古语问题集能够有效支持基于决策树的状态聚类与绑定方法.
Aiming at the current situation that Mongolian speech recognition system urgently needs to design a corresponding set of questions, the principles of Mongolian question set design, the selection of phonetic subsets, the completeness of phoneme lists, the classification of compound vowels, the concept of looseness and tightness Application and other key problems, a Mongolian standard pronunciation problem set is given.The results of several comparative experiments show that this Mongolian problem set can effectively support the state clustering and binding method based on decision tree.